I was chatting with a Business Development Director recently and he told me about his experience using ChatGPT. He’d watched his company’s video about AI and found it interesting. He’d used AI to alter a product image to answer a client’s question. Beyond that, he had little use for an LLM other than to ask the occasional research question.

This is a very typical exposure to LLMs. You watch a video or sit through a workshop. A couple of days later you can hardly remember what was said. You keep saying you’ll go back and watch the recording.

You do something you’ve never done before with ChatGPT. You think, "That's cool." But there is a limited economic benefit to it.

You try generating an email or short report and find it takes more time to edit than it would to write it yourself from scratch.

While you consider yourself an AI user, you can take it or leave it.

The UK's Department for Science, Innovation and Technology issued a report on AI adoption earlier this year. Only 16% of UK companies are using AI.

The primary uses are marketing and administration. The adopting companies report productivity gains, but few cite revenue increases as yet. With limited returns available, the companies restrict access to models a few layers below the latest and greatest versions.

This reported adoption is lower than that in surveys by Lloyds Banking Group and Microsoft UK. These focus on larger firms where adoption is over 50% and companies claim to be increasing profits. But the gains are self-reported and fewer than half of the average company’s workforce are regular users.

All of this is to be expected. Karl Popper said that knowledge cannot be poured from an expert into a novice like water between buckets. That’s why teaching techniques don’t work for large sections of the classroom.

Instead knowledge is acquired by curious minds. We lock onto what interests us and create an internal impression of how it works. The subsequent understanding may be similar, worse or better than that of whoever is explaining to us. It is never identical.

This means that companies spending money on AI videos and training workshops are ticking a box rather than enabling adoption. Staff will soon forget what they have been told. There is about to be much more of this.

Firms using AI in the EU must support AI literacy among staff, with financial penalties for not doing so from August 2, 2026.

Some firms do want to upskill their workforce, make the most of subscriptions they already pay for and reap the efficiency gains of eliminating manual work. This requires hands-on education and working with individuals over a period of weeks or months.

The organisations seeing the greatest return are treating AI adoption as professional development. They give people time to experiment, encourage them to solve problems from their own roles and provide someone who can answer questions as they arise. That is how occasional users become confident enough to rely on an LLM as part of their daily work.

Our experience at MSBC is that the level of help required is often too basic for tech companies. Not many people know how to structure a prompt to get effective answers. Few realise they can talk to an LLM much as they would brief a colleague. Very few know about setting up skills, which perform routine tasks that save hours a day.

This is not about replacing people. It is about alleviating the tedious aspects of most jobs. It is about freeing time for new projects. It is about making people feel better about their jobs and giving them greater agency at work. This is essential for improving results in any company.

MSBC has launched a dedicated company called Increments. It is an AI adoption partner rather than a training company. It is led by educators and backed by the tech knowledge of the parent company. Our aim is to get almost anyone to the point where they get genuine value from using an LLM every day.

We offer AI Fluency to get people started and AI Discovery for companies that want staff to use specific software such as Claude, ChatGPT and Microsoft Copilot.

We spend time figuring out how different people learn. We take time to understand their roles. We dedicate enough time for them to become confident with the tools.

The DSIT survey showed that there is a lot of superficial usage of AI. My Business Development Director is just one example. He needs to learn something different from his marketing, finance or HR department.

He will not do this with a one-size-fits-all video, or trial and error. He can get there with a dedicated AI adoption partner.